Cellular Distribution and Developmental Expression of AMP-Activated Protein Kinase Isoforms in Mouse Central Nervous System

Authors

  • Ann M. Turnley,

    1. The Walter and Eliza Hall Institute of Medical Research, Royal Melbourne Hospital, Victoria, Australia*St. Vincents Institute of Medical Research, Fitzroy, Victoria, AustraliaEndocrine-Metabolism Division, Dartmouth Medical School, Hanover, New Hampshire, U.S.A.
    Search for more papers by this author
  • David Stapleton,

    1. The Walter and Eliza Hall Institute of Medical Research, Royal Melbourne Hospital, Victoria, Australia*St. Vincents Institute of Medical Research, Fitzroy, Victoria, AustraliaEndocrine-Metabolism Division, Dartmouth Medical School, Hanover, New Hampshire, U.S.A.
    Search for more papers by this author
  • * Richard J. Mann,

    1. The Walter and Eliza Hall Institute of Medical Research, Royal Melbourne Hospital, Victoria, Australia*St. Vincents Institute of Medical Research, Fitzroy, Victoria, AustraliaEndocrine-Metabolism Division, Dartmouth Medical School, Hanover, New Hampshire, U.S.A.
    Search for more papers by this author
  • * Lee A. Witters,

    1. The Walter and Eliza Hall Institute of Medical Research, Royal Melbourne Hospital, Victoria, Australia*St. Vincents Institute of Medical Research, Fitzroy, Victoria, AustraliaEndocrine-Metabolism Division, Dartmouth Medical School, Hanover, New Hampshire, U.S.A.
    Search for more papers by this author
  • Bruce E. Kemp,

    1. The Walter and Eliza Hall Institute of Medical Research, Royal Melbourne Hospital, Victoria, Australia*St. Vincents Institute of Medical Research, Fitzroy, Victoria, AustraliaEndocrine-Metabolism Division, Dartmouth Medical School, Hanover, New Hampshire, U.S.A.
    Search for more papers by this author
  • and * Perry F. Bartlett

    1. The Walter and Eliza Hall Institute of Medical Research, Royal Melbourne Hospital, Victoria, Australia*St. Vincents Institute of Medical Research, Fitzroy, Victoria, AustraliaEndocrine-Metabolism Division, Dartmouth Medical School, Hanover, New Hampshire, U.S.A.
    Search for more papers by this author

  • Lippincott Williams & Wilkins, Inc., Philadelphia

  • The present address of Dr. D. Stapleton is Programme in Molecular Biology and Cancer, Samuel Lunenfeld Research Institute, 600 University Ave., Mount Sinai Hospital, Toronto, Ontario, M5G 1X5 Canada.

  • Abbreviations used: AMPK, AMP-activated protein kinase; DAPI, 4′,6-diamidino-2-phenylindole; E, embryonic day; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFAP, glial fibrillary acidic protein; P, postnatal day; PBS, phosphate-buffered saline; SDS, sodium dodecyl sulfate; SSC, saline—sodium citrate.

Address correspondence and reprint requests to Dr. A. M. Turnley at Neurobiology Laboratory, The Walter and Eliza Hall Institute of Medical Research, P.O. Royal Melbourne Hospital, Victoria 3050, Australia.

Abstract

Abstract: The mammalian AMP-activated protein kinase is a heterotrimeric serine/threonine protein kinase with multiple isoforms for each subunit (α, β, and γ) and is activated under conditions of metabolic stress. It is widely expressed in many tissues, including the brain, although its expression pattern throughout the CNS is unknown. We show that brain mRNA levels for the α2 and β2 subunits were increased between embryonic days 10 and 14, whereas expression of α1, β1, and γ1 subunits was consistent at all ages examined. Immunostaining revealed a mainly neuronal distribution of all isoforms. The α2 catalytic subunit was highly expressed in neurons and activated astrocytes, whereas the α1 catalytic subunit showed low expression in neuropil. The γ1 noncatalytic subunit was highly expressed by neurons, but not by astrocytes. Expression of the β1 and β2 noncatalytic subunits varied, but some neurons, such as granule cells of olfactory bulb, did not express detectable levels of either β isoform. Preferential nuclear localization of the α2, β1, and γ1 subunits suggests new functions of the AMP-activated protein kinase, and the different expression patterns and cellular localization between the two catalytic subunits α1 and α2 point to different physiological roles.

The mammalian AMP-activated protein kinase (AMPK) is a metabolic stress sensor/effector that becomes activated under conditions of nutrient starvation, vigorous exercise, or heat shock (Moore et al., 1991; Corton et al., 1994; Vavvas et al., 1997). Activation of AMPK is due to the rise in concentration of AMP, as well as phosphorylation by a kinase kinase (Davies et al., 1995). The AMPK phosphorylates a number of proteins involved in the regulation of cellular metabolism, including HMG-CoA reductase (Ferrer et al., 1985), acetyl-CoA carboxylase (Carling et al., 1987), hormone-sensitive lipase (Garton et al., 1989), and glycogen synthase (Carling and Hardie, 1989). Recently, Raf-1 (Sprenkle et al., 1997) and creatine kinase (Ponticos et al., 1998) have also been added to the growing list of AMPK substrates.

AMPK is a heterotrimeric protein kinase consisting of a catalytic α subunit and noncatalytic β and γ subunits (Mitchelhill et al., 1994; Stapleton et al., 1994). There are at least two isoforms for each subunit, termed α1 or α2 (Stapleton et al., 1996), β1 or β2 (Stapleton et al., 1997), and γ1 or γ2 (Gao et al., 1996). Homologues of AMPK have been reported in nonmammalian organisms, including rye, barley, tobacco, cauliflower, yeast, Drosophila, and C. elegans (Gao et al., 1996). The AMPK catalytic and noncatalytic domains share both structural and functional homology with the Saccharomyces cerevisiae protein kinase, Snf1 (Mitchelhill et al., 1994; Gao et al., 1995), indicating a high level of evolutionary conservation.

Although AMPK was isolated initially from liver, all three subunits are expressed by a wide variety of tissues, including lung, kidney, heart, skeletal muscle, and brain (Mitchelhill et al., 1994; Gao et al., 1995, 1996; Stapleton et al., 1996). We were interested in examining the expression of AMPK in brain, as it has an extremely high metabolic rate and high lipid content. The brain receives 15% of the body’s cardiac output, consumes 20% of the total oxygen, and yet comprises only 2% of total body weight (Chien, 1985). In view of the role of AMPK in both lipid and glucose metabolism, it was of interest to investigate the cellular distribution AMPK in brain. Preliminary reports of AMPK activity in cultured oligodendrocytes and astrocytes (Moore and Brophy, 1994; Cox et al., 1997) have been made. We have examined AMPK expression throughout development in the whole brain, as well as its cell-specific expression. We found that expression of the catalytic α2 and noncatalytic β2 subunits were up-regulated during embryonic days (E) 10-14, whereas α1, β1, and γ1 subunits were expressed consistently at all ages examined. Immunohistochemistry analysis using subunit-specific antibodies showed a mainly neuronal distribution, although the α2 and, to a lesser extent, the β2 subunits also showed astrocyte expression and were up-regulated in activated astrocytes.

MATERIALS AND METHODS

RNA extraction and northern hybridization

Brains were dissected from appropriately aged CBA mice and immediately frozen on dry ice. For the E10 time point, the neuroepithelium was used as the tissue source. Total RNA was then prepared from the frozen tissue using an RNeasy Mini Kit (Qiagen, Germany) according to the manufacturer’s instructions. The RNA (10 μg) from each sample was then electrophoresed on a 0.8% formaldehyde—agarose gel and capillary transferred to nylon membrane (Hybond N, Amersham, U.K.). The membranes were prehybridized for 2 h in 6× saline—sodium citrate (SSC), 5× Denhardt’s solution, 100 μg/ml herring sperm DNA, and 0.5% sodium dodecyl sulfate (SDS) at 68°C. Random-primed, 32P-labeled, full-length α1, α2, (Stapleton et al., 1996), β1, or γ1 AMPK (Gao et al., 1996) cDNA probes or the 3′ untranslated region of β2 cDNA was then added to the prehybridization solution at a specific activity of at least 106 cpm/ml and incubated overnight at 68°C. The membranes were then washed sequentially in 2× SSC/0.1% SDS at 68°C (2 × 15 min) and 0.2× SSC/0.1% SDS at 68°C (2 × 15 min) before autoradiography on Du Pont Reflection film with intensifying screens at -80°C for 3-72 h. The filters were sequentially stripped by boiling in 0.1% SDS for 10 min, and then reprobed with another AMPK cDNA and finally with a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) probe to check RNA loading levels. There was no cross-reactivity observed between the probes for different isoforms, as bands of different sizes were obtained for each probe. We also probed with γ2 cDNA, which was a NotI/EcoRI fragment of an EST (accession number R32671) obtained from the IMAGE Consortium, Washington University—Merck EST Project; however, we did not detect any γ2 expression by northern analysis.

Anti-AMPK antibodies and immunohistochemistry

Brains were removed from 12-, 18-, and 23-day-old CBA mice, or 16—18-day-old dysmyelinated transgenic mice (Turnley et al., 1991), embedded in OCT compound (Tissue Tek), and fresh-frozen on a dry ice/isopentane slurry. Sections (6 μm) were cut on a cryostat (Frigocut 2800E, Reichert Jung, Germany), transferred to 3-aminopropyltriethoxysilane (Sigma)-coated microscope slides, and air-dried for 15-30 min. The sections were then fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) for 10 min at room temperature, washed in PBS (3 × 5 min), and used immediately for immunohistochemical staining.

Anti-AMPK antibodies were raised in rabbits against peptides from the α1, α2, β1, β2, γ1, and γ2 AMPK subunits and affinity-purified, as previously described for α1 and α2 (Michell et al., 1996; Stapleton et al., 1996). The peptide sequences used for the preparation of the β and γ antibodies were as follows: for β1, amino acids 101-126 (CKLPLTRSQNNFVAILDLPEGEHQYK); for β2, amino acids 183-199 (CSPPGPYGQEMYVFRSE); for γ1, amino acids 319-331 (CQALVLTGGEKKP); and for γ2, the C-terminal sequence (CLTPAGAKQKETETE), deduced from the EST sequence (R32671). All antibodies detected their specific AMPK isoform by western blot analysis of partially purified AMPK fractions from rat liver. The β1 antibody showed slight cross-reactivity with the β2 isoform by western analysis, but did not appear to cross-react on tissue sections. All anti-AMPK antibodies were used at a dilution of 1:500 for immunostaining.

The neuronal nuclear marker, NeuN (Mullen et al., 1992), a mouse monoclonal antibody, was used to identify neurons. Rat anti-glial fibrillary acidic protein (GFAP), which detects only activated astrocytes, was purchased from Zymed Laboratories (San Francisco, CA, U.S.A.), whereas a rabbit anti-GFAP that detects most astrocytes was purchased from Dako (Denmark). Biotinylated anti-rabbit immunoglobulin, fluorescein-conjugated anti-mouse, anti-rabbit, or anti-rat immunoglobulin, and rhodamine-conjugated streptavidin were purchased from Southern Biotechnology (Birmingham, AL, U.S.A.).

For immunostaining, the sections were first blocked for 30 min in PBS containing 2% goat serum and 2% fetal calf serum. Primary antibodies in block were then added overnight at 4°C. After washing, secondary antibodies were added to the sections for 1 h at room temperature. Sections were again washed in PBS, and rhodamine-conjugated streptavidin was added for 30 min at room temperature, as well as the nuclear dye 4′, 6-diamidino-2-phenylindole (DAPI) (Sigma) at 1 μg/ml. After a final wash, sections were coverslipped in 1,4-diazabicyclo[2.2.2]octane (Merck). The use of animals was approved by the Royal Melbourne Hospital Animal Ethics Committee and performed in accordance with the principles of the National Health and Medical Research Council of Australia.

RESULTS

Developmental time course of AMPK mRNA expression in brain

Previous studies have shown that the AMPK subunits are expressed in adult brain (Gao et al., 1995, 1996; Stapleton et al., 1996). It was unknown when AMPK was expressed developmentally and whether expression was regulated. We therefore examined by northern analysis the developmental time course of AMPK expression to determine if there was any correlation of expression with differentiation of various neural cell types. RNA was made from mouse brains at various embryonic and postnatal ages, as indicated, and expression of the AMPK subunits was examined. The α1 catalytic subunit showed a fairly low, but consistent, level of expression at all ages, whereas expression of the α2 subunit showed two interesting features (Fig. 1A). The first was a dramatic increase in expression between E10 and E14, which correlated with the period of neuronal differentiation. The second was the presence of a much smaller α2 transcript, which comprised at least 50% of the α2 RNA at E10, and which gradually decreased until it was virtually undetectable by 8-12 days after birth. The non-catalytic β1 subunit was also present as two different sized transcripts, the larger of which was expressed at all ages, whereas the smaller had a similar time course of expression as the smaller α2 transcript. The β2 subunit, like α2, showed an increase in expression at E14 (Fig. 1B), whereas the γ1 subunit was expressed at similar levels at all ages examined. We could not detect expression of the γ2 subunit.

Figure 1.

Developmental time course of expression of AMPK isoform-specific RNAs in brain. Northern analysis of total RNA, prepared from brains of mice aged between E10 and postnatal day (P) 25, was performed using AMPK isoform-specific probes. A: AMPK α1 and γ1 were consistently expressed at all ages examined, whereas β1 showed a slight decrease in expression at later postnatal ages. AMPK α2 showed a dramatic increase in expression between E10 and E14. In addition, there were smaller RNA species detected by the α2 and β1 probes that were expressed between E10 and P8. B: AMPK β2 showed an increase in expression between E10 and E14, but was consistent thereafter. For both A and B, a GAPDH probe was used as a control for RNA loading.

FIG. 1.

AMPK α2 is the predominant catalytic subunit in brain and spinal cord

To determine the neural cell-specific expression of AMPK, brain cryostat sections were immunostained with antibodies to the various AMPK subunits. Sections were costained with antibodies to the neuronal nuclear marker, NeuN, which labels most neuronal cell types (Mullen et al., 1992), or to a form of GFAP expressed by activated astrocytes. Table 1 summarizes the staining data. As indicated by the northern analysis of AMPK RNA expression, the α2 catalytic subunit was expressed at higher levels than α1. AMPK α1 expression by immunostaining was barely detectable above background levels and was confined mainly to the neuropil. Whereas the soma of most neurons, including cortical neurons, pyramidal cells of the hippocampus (Fig. 2D), and cerebellar neurons (Fig. 3B), had low but detectable levels of α1 expression, the olfactory bulb neurons (Fig. 4B) and motor neurons (Fig. 5A) did not appear to express AMPK α1 much above background. The α2 subunit was expressed extensively throughout the brain, in what appeared to be a predominantly neuronal distribution. In gray matter areas, the α2 labeling was exclusively neuronal. Every AMPK α2-expressing cell coexpressed NeuN, whereas every NeuN-expressing neuron also expressed AMPK α2 (Fig. 2E). AMPK α2 was also expressed by neurons that are not detected by NeuN staining, such as the Purkinje cells of the cerebellum (Fig. 3C) and olfactory bulb mitral cells (Fig. 4C). The cerebellar granule cells also expressed the α2 subunit, but at lower levels than the Purkinje cells. Unlike α1, α2 was also highly expressed in most olfactory bulb neurons (Fig. 4C) and motor neurons (Fig. 5B).

Table 1. Cellular distribution and expression level of AMPK isoforms in CNS
CNS regionCell typeα1aα2 β1 β2 γ1 γ2
  1. -, not detectable above background; ±, barely detectable above background; +, low expression; ++, medium expression; +++, high expression; ++++, very high expression; ND, not determined.

  2. aα1 staining was confined mainly to low, diffuse staining of the neuropil and was not restricted to cell bodies.

  3. bγ2 staining of Bergmann glia, but not other astrocytes.

CortexNeuron++++- to ++++++++±
 Astrocyte------
Hippocampus/corpus callosumNeuron/pyramidal+++++ to +++++++++
 Astrocyte-- to +++- to +- to +--
 Activated astrocyteND++++- to ±± to ++-ND
CerebellumNeuron/granule+++±+++±
 Neuron/Purkinje++++± to +++++++++±
 Astrocyte-- to +++- to +- to +- ++b
Spinal cordMotor neuron±++± to +++±++++± to +
 Astrocyte-- to +++- to ++- to ++--
Facial nucleusMotor neuron- to ++++± to +++-++++± to +
Olfactory bulbNeuron/mitral-++++++-+++-
 Neuron/granule-+++--+++-
 Neuron/periglomular-++++++-++-
Figure 2.

Immunohistochemical analysis of AMPK expression in hippocampus/corpus callosum. Serial sagittal sections of P23 brain were immunostained for expression of the different AMPK isoforms and double-labeled with either anti-NeuN or anti-GFAP antibodies. A: A representative section stained for the neuronal nuclear marker NeuN. The locations of the pyramidal cells (pc) and the corpus callosum (cc) are indicated. B: A double-labeled section stained with rat anti-GFAP, which detects activated astrocytes. C: The same section as in B colabeled with rabbit anti-GFAP, which detects most astrocytes. This panel indicates the different expression patterns obtained with the two different anti-GFAP antibodies. D-L: AMPK immunostaining. D: AMPK α1 immunostaining was mainly in the neuropil. E: α2 expression was not only in neuropil but also in pyramidal cell nuclei (arrowhead), as well as cells of oligodendrocyte (left of asterisk) and astrocyte (arrow) morphology. F: In forebrain sections, from P16 dysmyelinated transgenic mice, in which there was a reactive astrocytosis, the number of astrocytes that expressed α2 and the level of α2 expression were increased, as indicated by the arrow. G: AMPK β1 was expressed in pyramidal cell nuclei at varying levels. H: Expression of β2 in pyramidal cells was cytoplasmic. I: In P16 transgenic, dysmyelinated mouse brain, the number of astrocytes (arrow) that expressed β2 was increased, but the level of expression was not increased to the same extent as α2. J: AMPK γ1 was expressed in pyramidal cells, as well as cells of oligodendrocyte distribution (arrow). K: γ2 was barely expressed above background. L: Normal rabbit serum gave a low background. Scale bar = 50 μm.

Figure 3.

Immunohistochemical analysis of AMPK expression in cerebellum. Serial sagittal sections of P30 cerebellum were immunostained for expression of the different AMPK isoforms. A: DAPI staining shows the localization of all cells, including granule cell nuclei (gc) and faint Purkinje cell nuclei (arrow). D: A representative section stained for NeuN shows the position of the granule cells, but does not label Purkinje cells. G: Background staining with normal rabbit serum was quite low. B: AMPK α1 expression was confined mainly to the neuropil. C:α2 was expressed in the nucleus of granule cells, Purkinje cells (arrow), stellate cells of the molecular layer (arrowhead), and neurons in the inferior colliculus (ic). E: AMPK β1 was expressed by few cells in the granular layer and at low levels in Purkinje cell nuclei (arrow), although it was expressed by some neurons in the inferior colliculus (ic). F:β2 expression was highest in the cytoplasm of Purkinje cells, including their dendrites (arrowhead), and was expressed at much lower levels in the cytoplasm of granule cells. H: AMPK γ1 was widely expressed in all neurons, like α2. I and K:γ2 expression was restricted mainly to Bergmann glia, with low expression in the neuropil. J: Distribution of GFAP-expressing astrocytes. The GFAP-positive vertical fibers of the Bergmann glia (arrow) are similar to the γ2-expressing vertical fibers in I and K (arrows). L: NeuN colabeling of the section in K indicates the granule cell layer. Scale bar in I = 100 μm. Scale bar in J and L = 50 μm.

Figure 4.

Immunohistochemical analysis of AMPK expression in olfactory bulb. Serial sections of P12 olfactory bulb were stained with antibodies to the different AMPK isoforms. A: DAPI staining shows the distribution of all cells, including granule cells (gc), mitral cells (arrow), and periglomerular cells (arrowhead). D: NeuN staining shows granule cells and some periglomerular neurons, but not mitral cells. G: Normal rabbit serum gave some background staining of mitral cells. B: AMPK α1 was not expressed much above background, except in glomeruli (arrow). C: AMPK α2 was extensively expressed in the nucleus of all neurons, with lower expression in the glomeruli. E: AMPK β1 expression was unusual, as it was expressed only by mitral cells and some periglomerular cells, but not granule cells. F:β2 did not appear to be expressed at all in olfactory bulb. H: AMPK γ1 showed the same widespread neuronal expression as α2, again with light glomerular staining. I:γ2 was not expressed above background. Scale bar = 100 μm.

Figure 5.

Immunohistochemical analysis of AMPK isoform expression in spinal cord. Serial transverse sections of P23 spinal cord were immunostained for the different AMPK isoforms. A: AMPK α1 was expressed in neuropil, including axons cut in cross-section (arrow), and at a low level in motor neurons (arrowhead). B: AMPK α2 showed nuclear expression in all neurons, as well as cytoplasmic expression in motor neurons (arrowhead) and astrocytes (arrow). C: AMPK β1 was highly expressed in the nucleus of motor neurons (arrowhead), but not interneurons, as well as in some astrocytes (arrow). D: AMPK β2 did not appear to be expressed by motor neurons, although it was expressed by some astrocytes. E: AMPK γ1 was expressed at extremely high levels in both the nucleus and cytoplasm of motor neurons, with prominent axonal staining of both motor neurons (arrow) and transversely cut white matter axons (arrowhead). It was not expressed by astrocytes. F: AMPK γ2 was expressed at low levels in motor neurons. G: Rat anti-GFAP colabeling of the section in C shows staining of activated astrocytes. The arrow points to a GFAP-positive astrocyte fiber that also expressed AMPK β1 (arrow in C), and that had a similar expression pattern to the astrocyte staining in B and D. H: DAPI staining of the section shown in B indicates the presence of a larger number of cells than were labeled with the anti-AMPK antibodies. The arrowhead denotes the same motor neuron as labeled in B. Scale bar = 50 μm.

TABLE 1.

FIG. 2.

FIG. 3.

FIG. 4.

FIG. 5.

In myelinated white matter areas, such as corpus callosum or cerebellar white matter, scattered cells, which did not label with the neuronal marker NeuN, were detected by anti-α2 (Fig. 2E). Some of these cells had the characteristic localization of intrafascicular oligodendrocytes, but most were of astrocyte morphology and were more prominent in spinal cord white matter (Fig. 5B). Astrocyte labeling was confirmed by use of a rat anti-GFAP antibody that only detects activated astrocytes. This antibody detected a small number of astrocytes in white matter tracts, which colabeled with the anti-α2 antibody (Figs. 2E and 5B). The rat anti-GFAP antibody did not detect all astrocytes in white matter and did not detect astrocytes in gray matter (Fig. 2B). A rabbit anti-GFAP antibody, which detects most astrocytes, was used to examine the overall astrocyte distribution. This second anti-GFAP antibody detected a larger number of astrocytes in white matter than detected by α2 immunostaining, as well as astrocytes in the gray matter (Fig. 2C); however, it could not easily be used for colabeling with anti-AMPK antibodies, as all were raised in rabbits.

Expression of noncatalytic AMPK subunits in brain and spinal cord

Varied expression patterns for the noncatalytic AMPK subunits were observed. The AMPK γ1 subunit had a similar cellular distribution and level of expression as the α2 catalytic subunit. Expression of γ1 was largely confined to neurons (Figs. 2J, 3H, and 4H). Motor neurons in the spinal cord (Fig. 5E) and facial motor nucleus (data not shown) expressed γ1 at extremely high levels. However, γ1 was rarely, if at all, expressed in astrocytes (Figs. 2J and 5E). Some cells in white matter tracts with the distribution of oligodendrocytes also expressed the γ1 isoform (Fig. 2J). The antibody used to immunostain the γ2 isoform detected very little, if any, expression in all areas of the brain examined (Figs. 2K, 4I, and 5F) except for vertical fibers in the molecular layer of the cerebelium (Fig. 3I and K). Other sections, immunostained for GFAP, showed similar vertical fibers in the molecular layer, raising the possibility that the γ2-positive fibers were Bergmann glia (Fig. 3J).

The β subunits also showed a differential expression pattern. Although both were expressed predominantly in neurons, there were significant, discrete differences. The β1 subunit was not expressed by all neurons. In each area examined, and especially noticeable in cortex, a subpopulation of neurons did not express β1, and of those that expressed it, the level of expression varied from low to high (Figs. 2G, 3E, and 5C). In some areas, such as the granule cell layer of the olfactory bulb, entire neuronal populations did not express β1 (Fig. 4E). The β2 subunit was expressed at low to medium levels by most neurons (Fig. 2H); however, Purkinje cells of the cerebellum expressed it at extremely high levels (Fig. 3F). Both β2 and, to a lesser extent, β1 were also expressed in a subpopulation of white matter astrocytes and, like α2, showed a higher level of astrocyte expression in spinal cord white matter (Fig. 5C and D).

AMPK α2, β1, and γ1 are localized to the nucleus in neurons

In addition to differences in cellular distribution of the different AMPK subunits, it was also evident that the intracellular distribution varied. In all areas of the CNS examined, the neuronal expression of the catalytic α2 subunit was preferentially localized to the nucleus. The neuronal marker NeuN was also localized to the nucleus, although some cytoplasmic expression was observed, as reported (Mullen et al., 1992). The nuclear dye DAPI, NeuN, and AMPK α2 immunostaining were coincident in all neurons detected by NeuN. Nuclear localization of α2 staining of neurons not detected by NeuN, such as Purkinje cells in the cerebellum and mitral cells in the olfactory bulb, was also observed and confirmed by the use of DAPI (Figs. 2E, 3C, and 4C). Cytoplasmic expression of AMPK α2 could be detected in many cases, but the nuclear expression was always higher. This was easily visualized in spinal cord motor neurons, which have a large cytoplasm, as well as a large nucleus (Fig. 5B). Although the α2 subunit was localized to neuronal nuclei, most α2 expression in astrocytes was cytoplasmic. Only a few astrocytes showed both nuclear and cytoplasmic expression, which were of similar intensities, although cultured primary astrocytes showed increased nuclear localization of α2 (data not shown). Expression of the α1 catalytic subunit, where detectable, was mainly, if not totally, cytoplasmic. Expression of α1 could sometimes be detected in axons, such as axons cut in transverse section in spinal cord white matter (Fig. 5A), whereas in other areas, a diffuse staining of the neuropil was observed, such as in the glomeruli of the olfactory bulb (Fig. 4B).

The noncatalytic β1 and γ1 subunits not only had a cellular distribution similar to that of α2, but were also similarly preferentially localized to the nucleus (Figs. 2G and J, 3E and H, 4E and H, and 5C and E). One difference was in motor neurons, where γ1 expression was extremely high not only in the nucleus, but also in the cytoplasm (Fig. 5E). In other neurons, although γ1 cytoplasmic expression could be quite high, it usually was not as high as the nuclear expression. Of the other two noncatalytic subunits examined, β2 was exclusively localized to the cytoplasm, and γ2, where it could be detected, also appeared to be cytoplasmic (Figs. 2H and 3F, I, and K). Cytoplasmic localization of β2 is particularly evident in Purkinje cells, where not only the cell soma, but also the dendrites, were extensively stained (Fig. 3F). Table 2 provides a summary of nuclear versus cytoplasmic localization of the various AMPK isoforms.

Table 2. Nuclear versus cytoplasmic localization of the AMPK isoforms
Cell typeα1 α2 β1 β2 γ1 γ2
  1. c, cytoplasmic; n, nuclear; >, greater than; >>, much greater than; ≥, greater than or equal to; NA, not applicable.

  2. aα1 staining was confined mainly to low, diffuse staining of the neuropil and was not restricted to cell bodies.

  3. bγ2 staining of Bergmann glia, but not other astrocytes.

Neurons c >> nan >> cn >> ccn ≥ cc > n
Astrocytes—white matter NAc > nccNA cb

TABLE 2.

AMPK expression is up-regulated in activated astrocytes

A small number of apparently activated astrocytes in normal mice expressed the α2 AMPK, suggesting that activated astrocytes may show increased levels of AMPK expression. To examine this, we analyzed AMPK expression in brains of dysmyelinating transgenic mice that overexpressed class I major histocompatibility complex molecules in oligodendrocytes (Turnley et al., 1991). These mice exhibited an astrocytic gliosis in white matter tracts. Colabeling with anti-AMPK antibodies and the anti-activated astrocyte GFAP antibody showed that all cells with astrocyte morphology that expressed α2 AMPK also expressed GFAP. In addition, there was an increase in both the number of astrocytes that expressed α2 AMPK and the level of expression (Fig. 2F). The β2 AMPK subunit also showed a slightly increased level of expression in the astrocytes of the dysmyelinated mice; however, the increase was not as great as that observed for α2 expression (Fig. 2I). Expression of the other AMPK subunits did not appear to be increased in activated astrocytes.

DISCUSSION

We have shown that each of the AMPK subunits is expressed in the CNS from at least the time of neuroepithelial formation until the adult. By immunostaining, expression was shown to vary depending on the region of the CNS and was mainly neuronal, although activated astrocytes also expressed the α2 and β2 isoforms.

Analysis of AMPK RNA expression in brain

The AMPK α2 subunit was found to be the most highly expressed of the catalytic AMPK subunits and, like the β2 subunit, showed regulation of RNA expression, with a dramatic increase from E10 to E14. RNA splice variants were present for α2, β1, and γ1, which for α2 and β1 were not detectable after P8. Of these splice variants, the α2 appeared to be the most significant, comprising at least 50% of the AMPK α2 message at E10. This had been reported previously in multiple tissues (Gao et al., 1995) using a 3′ untranslated region probe, but its sequence and function are unknown.

The developmental time course of expression suggests that α2 and β2 may play a specific role in neuronal cell function, being dramatically up-regulated during the major time of neurogenesis (Abney et al., 1981). The presence of the other subunits at similar levels at all ages examined may suggest more general roles. Although α2 was also shown to be expressed by astrocytes, there was no apparent large increase in expression at E16, which is the time of commencement of astrocyte differentiation (Abney et al., 1981). Immunostaining showed that astrocyte expression of α2, especially in brain, was confined mainly to reactive white matter astrocytes, which are present in low numbers in normal brain. Therefore, they would not make much contribution to the level of RNA expressed when whole brain was examined, as the number of neurons expressing α2 was much greater. We could not detect RNA expression of the γ2 isoform. This may be expected, however, given the extremely small number of cells in cerebellum that were shown to express it by immunohistochemistry, whereas our RNA was isolated from whole brain.

Cellular distribution of AMPK immunostaining in the CNS

Immunostaining for the various AMPK isoforms correlated with the northern analysis of expression. The α1 isoform was widely distributed, and although there was some neuronal localization, in most areas staining was more diffuse and corresponded to the neuropil. The high level of α2 RNA expression was reflected in the high level of immunostaining, although γ1 had similar high levels of expression by immunostaining, but lower RNA levels. Previously, we reported a lack of tight correlation between RNA levels of AMPK subunits in different tissues (Gao et al., 1995). This may reflect different antibody affinities or RNA stability. The γ1 and β1 isoforms had a mainly neuronal distribution, although their RNA was expressed at E10 at similar levels to later stages, when there were very few mature neurons. This suggested that γ1 and β1 were also expressed in neural precursor cells, but remains to be examined. At present, the developmental time course of expression of γ2 is unknown. Given the different time courses of expression of the AMPK isoforms throughout development, it would be interesting to know if the cell-specific expression of the AMPK isoforms changes over time. The results described in Table 1 only reflect immunostaining of CNS from young adult mice. Immunostaining of CNS at different stages of development for AMPK isoform localization is yet to be performed.

The only AMPK subunits we detected in astrocytes were the α2 catalytic subunit and the β2 and, to a lesser extent, β1 noncatalytic subunits. We have tested two known γ subunits, which did not seem to be expressed in astrocytes, with the exception of γ2 in Bergmann glia. It is possible that other, as yet unidentified, γ subunits may be expressed in these astrocytes.

Up-regulation of AMPK expression in activated astrocytes

The only astrocytes that expressed AMPK in normal brain were a population of white matter astrocytes, which also stained with an antibody that detects GFAP of reactive, but not normal, astrocytes. Injury to the CNS results in an astrocytic gliosis, whereby quiescent astrocytes become reactive. Expression of GFAP is up-regulated, and different epitopes become accessible to some GFAP antibodies (for review, see Wu and Schwartz, 1998). The astrocytes become enlarged, migrate to the site of injury, and release a variety of cytokines and growth factors (Ridet et al., 1997). AMPK activity in astrocytes has been reported previously (Cox et al., 1997). The astrocytes in this study were in culture for 28 days and so would have many of the characteristics of activated astrocytes, e.g., up-regulation of GFAP. We showed here that most astrocytes do not normally express AMPK, but that expression can be increased when there is an increase in metabolic activity, such as during reactive gliosis. The subunit expression in activated astrocytes included α2 and β2, but the γ subunit isoform is not known.

Nuclear versus cytoplasmic localization of the AMPK subunits

One surprising finding was the differential localization of the AMPK isoforms between the nucleus and cytoplasm of neurons. There are no known nuclear localization signals in the AMPK α2, β1, or γ1 subunits. However, several other kinases are translocated to the nucleus in the absence of known nuclear localization signals, such as protein kinase C (Schmalz et al., 1998) and mitogen-activated protein kinase (Fukuda et al., 1997). It may be possible for the 40-kDa β1 and 38-kDa γ1 AMPK subunits to enter the nucleus by passive diffusion through the nuclear pore, but the 63-kDa α2 isoform would probably be too large to enter the nucleus by such a mechanism (Gorlich and Mattaj, 1996; Nigg, 1997).

The role of nuclear localized AMPK α2 is not yet known. We have also found α2 nuclear localization in the liver and skeletal muscle (data not shown), suggesting it is not a brain-specific phenomenon. Two recent reports suggest that the AMPK may play a role in inhibiting glucose-activated gene expression in hepatocytes (Foretz et al., 1998; Leclerc et al., 1998).

Potential roles of AMPK in brain

The level of expression of the various AMPK isoforms varied in different regions of the brain; these data are summarized in Table 1. The AMPK isoforms may simply be regulating normal metabolism and energy usage; however, they may also have some more specialized functions within the nervous system. Not only do the α1 and α2 isoforms show slight differences in their substrate preferences (Michell et al., 1996; Woods et al., 1996a), we show here that they are differentially expressed in the nervous system, which may suggest differential physiological roles for each isoform.

Although most neurons expressed high levels of the catalytic α2 subunit and the associated noncatalytic γ1 and β1 or β2 subunits, the level of expression appeared to be highest in areas of brain that have been reported to have the highest glucose utilization. Neurons, in general, have a very high metabolic rate and high glucose usage; however, local glucose utilization varies in different regions of the brain. Neural structures reported to have very high levels of glucose utilization include Purkinje cells of the cerebellum and pyramidal cells of the hippocampus (Pertsch et al., 1988), areas in which we found very high levels of AMPK expression. Indeed, Purkinje cells were the only cell type in which we found the β2 subunit expressed at extremely high levels, in addition to lower levels of the other subunits. The high level of α2 expression, coupled with a similarly high level of β2 and γ2 expression in the same defined population of high glucose-using cells, may suggest a more specialized role for an α2/β2/γ1 or an α2/β1/γ1 heterotrimer in glucose metabolism. This may also be the case with highly metabolically active astrocytes. The nuclear localization of β1 and the cytoplasmic localization of β2 may suggest different regulatory mechanisms or different substrate specificities.

Given the more diffuse staining pattern of α1 and its localization in the neuropil, it is possible that α1 is involved in the regulation of membrane excitability in neurons, especially during metabolic stress. ATP synthesis and metabolism have been localized to neurites (Tolkovsky and Suidan, 1987), and metabolic stresses, such as oxygen or glucose deprivation and membrane depolarization, caused local ATP breakdown in neurons (Tolkovsky and Suidan, 1987; Rego et al., 1997). During metabolic stress, ATP-modulated K+ channels play an important role in modulating membrane excitability, with different ATP concentrations modulating different K+ channels (Jiang et al., 1994; Jiang and Haddad, 1997). In addition, ATP modulates Ca2+-dependent currents in metabolically stressed neurons (Stapleton et al., 1995). As AMPK is activated by decreased levels of ATP (Corton et al., 1994), it seems reasonable that it functions to monitor local metabolic stress and ATP depletion in neurites.

AMPK isoform association and regulation of activity

The regulation of potentially functional AMPK activity appears to be different in neurons and astrocytes. Neurons show widespread expression of both α isoforms, with α1 mainly in the neuropil and α2 mainly in the nucleus, and with γ1 and β isoforms present in both cytoplasm and nucleus. Significant AMPK activity is obtained only when all three subunits are present (Dyck et al., 1996; Woods et al., 1996b), and in vitro binding studies have shown that the β1 subunit is required to stabilize the α2 and γ1 interaction (Woods et al., 1996b). In astrocytes, however, whereas α2 and, to a lesser extent, β2 and β1 were expressed, expression of γ1 was not detected. This may indicate that astrocytes express an unknown γ isoform.

Note added in proof: Preferential nuclear localization of α2 AMPK was recently shown by Salt et al. (1998).

Acknowledgements

We would like to thank Cheryl Augustine and Frosa Katsis for expert technical assistance. A.M.T. is an Australian Research Council postdoctoral fellow. This work is supported in part by the National Health and Medical Research Council of Australia and by an NIH grant (DK35712) to L.A.W. B.E.K. is a National Health and Medical Research Council Fellow.

Ancillary